Aktualności14 października 2026 LLMs don't reason: Graepel wants an explicit epistemic state
Thore Graepel of UCL, formerly on the AlphaGo team, argues language models do not reason because they lack a persistent epistemic state separate from knowledge.
